Skip to content

Docx

DOCXConverter

Bases: BaseConverter

A component for converting DOCX files to Documents using the python-docx library.

Initializes the object with the configuration for converting documents using python-docx.

Parameters:

Name Type Description Default
document_creation_mode Literal['one-doc-per-file', 'one-doc-per-page']

Determines how to create Documents from the elements of the Word document. Options are: - "one-doc-per-file": Creates one Document per file. All elements are concatenated into one text field. - "one-doc-per-page": Creates one Document per page. All elements on a page are concatenated into one text field. Defaults to "one-doc-per-file".

required
Usage example
from dynamiq.components.converters.docx import DOCXConverter

converter = DOCXConverter()
documents = converter.run(paths=["a/file/path.docx", "a/directory/path"])["documents"]
Source code in dynamiq/components/converters/docx.py
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
class DOCXConverter(BaseConverter):
    """
    A component for converting DOCX files to Documents using the python-docx library.

    Initializes the object with the configuration for converting documents using
    python-docx.

    Args:
        document_creation_mode (Literal["one-doc-per-file", "one-doc-per-page"], optional):
            Determines how to create Documents from the elements of the Word document. Options are:
            - `"one-doc-per-file"`: Creates one Document per file.
                All elements are concatenated into one text field.
            - `"one-doc-per-page"`: Creates one Document per page.
                All elements on a page are concatenated into one text field.
            Defaults to `"one-doc-per-file"`.

    Usage example:
        ```python
        from dynamiq.components.converters.docx import DOCXConverter

        converter = DOCXConverter()
        documents = converter.run(paths=["a/file/path.docx", "a/directory/path"])["documents"]
        ```
    """

    document_creation_mode: Literal[DocumentCreationMode.ONE_DOC_PER_FILE, DocumentCreationMode.ONE_DOC_PER_PAGE] = (
        DocumentCreationMode.ONE_DOC_PER_FILE
    )

    xml_key: str = "{http://schemas.openxmlformats.org/wordprocessingml/2006/main}val"
    xml_namespaces: dict[str, str] = {
        "w": "http://schemas.microsoft.com/office/word/2003/wordml",
        "a": "http://schemas.openxmlformats.org/drawingml/2006/main",
        "r": "http://schemas.openxmlformats.org/officeDocument/2006/relationships",
    }
    image_placeholder: str = "<!-- image -->"
    blip_tag: str = "{http://schemas.openxmlformats.org/drawingml/2006/main}blip"

    def _process_file(self, file: Path | BytesIO, metadata: dict[str, Any]) -> list[Any]:
        """
        Process a single Word document and create documents.

        Args:
            file (Union[Path, BytesIO]): The file to process.
            metadata (Dict[str, Any]): Metadata to attach to the documents.

        Returns:
            List[Any]: A list of created documents.

        Raises:
            TypeError: If the file argument is neither a Path nor a BytesIO object.
        """
        if isinstance(file, Path):
            with open(file, "rb") as upload_file:
                file_content = BytesIO(upload_file.read())
                file_path = upload_file.name
        elif isinstance(file, BytesIO):
            file_path = get_filename_for_bytesio(file)
            file_content = file
        else:
            raise TypeError("Expected a Path object or a BytesIO object.")

        file_content.seek(0)
        elements = DocxDocument(file_content)
        return self._create_documents(
            filepath=file_path,
            elements=elements,
            document_creation_mode=self.document_creation_mode,
            metadata=metadata,
        )

    def _create_documents(
        self,
        filepath: str,
        elements: DocxDocument,
        document_creation_mode: DocumentCreationMode,
        metadata: dict[str, Any],
        **kwargs,
    ) -> list[Document]:
        """
        Create Documents from the elements of the Word document.
        """
        docs = []
        if document_creation_mode == DocumentCreationMode.ONE_DOC_PER_FILE:
            text_content = []

            for element in elements.element.body:
                tag_name = element.tag.split("}", 1)[-1] if "}" in element.tag else element.tag

                # Check for inline images
                contains_blip = any(child.tag == self.blip_tag for child in element.iter())

                if element.tag.endswith("p"):
                    text = self._process_paragraph(element, elements)
                    if text.strip():
                        text_content.append(text)
                elif element.tag.endswith("tbl"):
                    text = self._process_table(element, elements)
                    if text.strip():
                        text_content.append(text)
                elif contains_blip:
                    text_content.append(self.image_placeholder)
                elif tag_name in ["sdt"]:
                    sdt_content = element.find(".//w:sdtContent", namespaces=self.xml_namespaces)
                    if sdt_content is not None:
                        paragraphs = sdt_content.findall(".//w:p", namespaces=self.xml_namespaces)
                        for p in paragraphs:
                            text = self._process_paragraph(p, elements)
                            if text.strip():
                                text_content.append(text)

            full_text = "\n\n".join(text_content)

            metadata = copy.deepcopy(metadata)
            metadata["file_path"] = filepath
            docs = [Document(content=full_text, metadata=metadata)]

        elif document_creation_mode == DocumentCreationMode.ONE_DOC_PER_PAGE:
            sections = self._split_into_sections(elements)
            metadata = copy.deepcopy(metadata)
            metadata["file_path"] = filepath

            for idx, section_content in enumerate(sections, start=1):
                section_metadata = copy.deepcopy(metadata)
                section_metadata["page_number"] = idx
                docs.append(Document(content=section_content, metadata=section_metadata))

        return docs

    def _process_paragraph(self, paragraph_element, docx_obj) -> str:
        """
        Process a paragraph element with formatting and hyperlinks.
        """
        paragraph = Paragraph(paragraph_element, docx_obj)
        text = paragraph.text

        paragraph_elements = self._get_paragraph_elements(paragraph)
        is_list_item, list_marker, list_level = self._check_for_list_item(paragraph)
        is_header, header_level = self._check_for_header(paragraph)

        formatted_text = ""

        if is_header:
            header_prefix = "#" * header_level
            formatted_text = f"{header_prefix} {text}"
        elif is_list_item:
            indent = "  " * (list_level - 1) if list_level > 1 else ""
            formatted_text = f"{indent}{list_marker} {text}"
        else:
            formatted_parts = []
            for txt, format_info, hyperlink in paragraph_elements:
                if not txt.strip():
                    continue

                if format_info:
                    if format_info.get("bold", False):
                        txt = f"**{txt}**"
                    if format_info.get("italic", False):
                        txt = f"*{txt}*"
                    if format_info.get("underline", False):
                        txt = f"_{txt}_"

                if hyperlink:
                    txt = f"[{txt}]({hyperlink})"

                formatted_parts.append(txt)

            formatted_text = " ".join(formatted_parts)

        return formatted_text

    def _process_table(self, table_element, docx_obj) -> str:
        """
        Process a table element with enhanced features like cell spanning and formatting.
        """
        table = Table(table_element, docx_obj)

        # Check for single-cell tables
        if len(table.rows) == 1 and len(table.columns) == 1:
            cell_text = table.rows[0].cells[0].text
            if cell_text.strip():
                return cell_text

        table_rows = []

        if table.rows:
            header_cells = []
            for cell in table.rows[0].cells:
                header_cells.append(cell.text.strip() or "")
            table_rows.append("| " + " | ".join(header_cells) + " |")
            table_rows.append("| " + " | ".join(["---"] * len(header_cells)) + " |")

        cell_set = set()
        for row_idx, row in enumerate(table.rows):
            # Skip the header row
            if row_idx == 0:
                continue

            row_cells = []
            for cell in row.cells:
                if cell._tc in cell_set:
                    continue
                cell_set.add(cell._tc)

                cell_text = cell.text.strip() or ""
                row_cells.append(cell_text)

            if row_cells:
                table_rows.append("| " + " | ".join(row_cells) + " |")

        return "\n".join(table_rows)

    def _get_paragraph_elements(self, paragraph):
        """
        Extract paragraph elements (with the formatting and hyperlinks).
        """
        if paragraph.text.strip() == "":
            return [("", None, None)]

        paragraph_elements = []

        for content in paragraph.iter_inner_content():
            if isinstance(content, Hyperlink):
                text = content.text
                hyperlink = content.address if hasattr(content, "address") else None
                format_info = self._get_format_from_run(content.runs[0] if content.runs else None)
                if text.strip():
                    paragraph_elements.append((text, format_info, hyperlink))
            elif isinstance(content, Run):
                text = content.text
                format_info = self._get_format_from_run(content)
                if text.strip():
                    paragraph_elements.append((text, format_info, None))

        if not paragraph_elements and paragraph.text.strip():
            paragraph_elements.append((paragraph.text.strip(), None, None))

        return paragraph_elements

    def _get_format_from_run(self, run):
        """
        Extract formatting information from a run.
        """
        if not run:
            return None

        format_info = {}
        if hasattr(run, "bold") and run.bold:
            format_info["bold"] = True
        if hasattr(run, "italic") and run.italic:
            format_info["italic"] = True
        if hasattr(run, "underline") and run.underline:
            format_info["underline"] = True

        return format_info if format_info else None

    def _check_for_list_item(self, paragraph):
        """
        Check if a paragraph is a list item and return relevant information.
        """
        numbering_properties = paragraph._element.find(".//w:numPr", namespaces=paragraph._element.nsmap)
        if numbering_properties is not None:
            numbering_id_elem = numbering_properties.find("w:numId", namespaces=paragraph._element.nsmap)
            indexing_level_elem = numbering_properties.find("w:ilvl", namespaces=paragraph._element.nsmap)

            numbering_id = numbering_id_elem.get(self.xml_key) if numbering_id_elem is not None else None
            indexing_level = indexing_level_elem.get(self.xml_key) if indexing_level_elem is not None else None

            if numbering_id and numbering_id != "0":
                level = int(indexing_level) + 1 if indexing_level else 1
                # Check if the paragraph is a numbered list item or a bullet list item
                is_numbered = "Number" in paragraph.style.name if paragraph.style and paragraph.style.name else False
                marker = f"{level}." if is_numbered else "•"
                return True, marker, level

        return False, "", 0

    def _check_for_header(self, paragraph):
        """
        Check if a paragraph is a header and return the level.
        """
        if not paragraph.style:
            return False, 0

        style_id = paragraph.style.name.lower() if paragraph.style.name else ""

        heading_string = "heading"
        heading_pattern = r"heading\s*(\d+)"

        if heading_string in style_id:
            match = re.search(heading_pattern, style_id)
            if match:
                level = int(match.group(1))
                return True, level

        return False, 0

    def _split_into_sections(self, doc: DocxDocument) -> list[str]:
        """
        Split the document into sections based on section breaks.
        """
        sections = []
        current_section = []

        for element in doc.element.body:
            if element.tag.endswith("sectPr"):
                if current_section:
                    sections.append("\n\n".join(current_section))
                    current_section = []
            elif element.tag.endswith("p"):
                text = self._process_paragraph(element, doc)
                if text.strip():
                    current_section.append(text)
            elif element.tag.endswith("tbl"):
                text = self._process_table(element, doc)
                if text.strip():
                    current_section.append(text)

        if current_section:
            sections.append("\n\n".join(current_section))
        if not sections:
            return [""]

        return sections